As the season continues to gain steam into the summer months, the news becomes further populated with pitchers succumbing to ulnar collateral ligament (UCL) tears, requiring Tommy John surgery (TJS), with some of them succumbing to their second tear. There have been 23 pitchers who have had their 2014 seasons shelved due to requiring surgery and with the rising numbers headed to the disabled list with UCL tears, a concomitant increase in the questions surrounding the why and how of their injury naturally comes. As previously discussed here at Beyond the Box Score, the reasons behind the etiology of UCL tears and their purported increase in incidence this season are complex; in some instances, certain variables that might play a role in tears (and re-tears) may never be publicly acknowledged or discussed.
Undaunted by the complexity or occasional lack of data, let's again revisit the UCL tear phenomenon, in particular, the space between those who have had a UCL tear and those who have had two—beyond the previously alluded to leverage and tobacco use, are there other forms of available data that could potentially explain why certain players suffer more than one tear?
When it comes to publicly available data on pitchers, our bread and butter remains PITCHf/x data and with it, we are able to look at a numbers of pitch variables, including release points, velocities, location, and movement for pitchers and their various pitch types. Previous work has shown that TJS pitchers tend to throw a little harder across the board compared to their non-TJS cohorts. Also, a consistent release point across all types of pitches thrown is a crucial aspect of pitching performance; it is also hinted at playing a role in the long-term health of a pitcher. Could release point variations between pitch repertoire before and after Tommy John surgery help explain some of the differences between one- and two-TJS pitchers? Could there be other aspects hinted at in the PITCHf/x data that makes a pitcher more apt to suffer a re-tear?
In my previous pieces on TJS, the 'dataset' comprised of Stephen Strasburg and Kris Medlen, who represented the one-TJs and two-TJS populations, respectively. Let's try to expand our N, with a criteria in mind. Beyond being a member of either the one- or two-TJS Club, a pitcher should also have at least 50 innings pitched before and after their TJS, have pitched their entire careers in the PITCHf/x age (from 2007 to current), and to keep tabs on any confounding effects of increased workloads and injury potentials in mind, be under 30 years of age. It isn't a perfect matching criteria, but it will do for this exploratory dataset. With this criteria, it was hoped to get at least three of each group; unfortunately, even with relaxing the innings criteria to 30 innings, we only come up with four pitchers, two in each of the one-TJS and two-TJS categories: Brett Anderson (one), Brandon Beachy (two), Kris Medlen (two), and Stephen Strasburg (one).
We have our sample, now let's put their data to the test, in the form of Stephen Loftus' Similarity Scores. I won't go into great detail on the intricacies of his algorithm—you can find the nitty gritty here and here—but briefly, the application of this algorithm on these pitchers' PITCHf/x data will allow us to see how similar or different of a pitcher they were before and after TJS, using pitch velocity, movement, release point, location, batter handedness, and pitch sequencing as measurements. Score interpretation of how similar/different a pitcher is pre- to post-TJS is as follows:
|0.8 - 1.0||Extremely Similar|
|0.6 - 0.8||Reasonably Similar|
|0.5 - 0.6||Somewhat Similar|
|0.4 - 0.5||More Different than Similar|
|0.2 - 0.4||Mostly Different|
|0.0 - 0.2||Entirely Different|
...and here is how each pitcher fared against themselves, pre- and post-TJS:
Pitch Sim is the overall similarity score for each pitcher. It is followed by additional scores for each pitch in a pitcher's repertore: CHSim denotes changeup, CUSim is for curveball, FFSim is for four-seam fastball, FTSim for two-seam fastball, and SLSim denotes slider.
Overall, we don't see much difference before and after TJS for any of our pitchers, with each of them remaining in the 'extremely similar' category after UCL tear and recovery. It's in the pitch-specific data that we begin to see some interesting deviations from pre-surgery performances—Anderson is a particularly interesting case, especially when looking at how different some of his offspeed pitches have become and are used post-TJS. While we can't make any dramatic inferences with such a small sample, the less similar fastballs seen in the two-TJS duo could be an interesting direction to go in digging deeper into explaining why they fell victim to another UCL tear, and parsing out what particular variables are responsible for driving the differences seen in pre- and post-TJS fastball between our two groups.
A big thank you to Stephen Loftus for his assistance with this article.
PITCHf/x data courtesy of MLBAM.
Stuart Wallace is an associate managing editor and writer at Beyond The Box Score. You can follow him on Twitter at @TClippardsSpecs.
Stephen Loftus is a featured writer at Beyond The Box Score. You can follow him on Twitter at @stephen__loftus.